I get AI working inside the systems you already run.
I'll also tell you where it doesn't belong.
The short version
AI projects depend on the data and systems you already run. That's integration work, and I've done it for ten years: ERP migrations, telecom platforms, public-sector automation, and now LLMs in production alongside all of it. You get a senior architect who owns the work from scoping to handover, and leaves your team with something they can run without me.
Who I work with
Three kinds of teams bring me in, each for a different reason.
When a client project needs one more senior person. I work under your brand, through my own EU company, and hand over cleanly when the work is done.
When an AI or integration workstream needs someone to own it. I work to a defined scope and end date, and leave documentation your team can carry on with.
When the product depends on AI or integrations. I've been the founding engineer at three startups: I build the first version, make the trade-offs that get it to market, and help hire the team that takes over. I can also work directly with your customers.
What I do
Picking the use cases where AI pays off, connecting models to your data and workflows, and testing them against your own data before anyone depends on them. Sometimes that test shows a cheaper model is good enough, or that plain code does the job better.
I design APIs and partner integrations that run in production. I've put eight telecom carriers behind one API, and moved dozens of ERP and EDI integrations to a new platform while the business kept running on them.
I join discovery, scoping and pre-sales calls, and onboard partners. I turn what a client asks for into a scope and an architecture, and explain the trade-offs so the non-technical people can make the decision.
Selected work
I wrote the winning tender and led the technical work through to handover. The automation engine cut case handling from about an hour to under a minute, with a full audit trail.
We tested GPT-4 as the decision engine. It wasn't consistent enough for regulatory decisions, so we built dedicated detection logic instead.
I led the move of ERP, EDI and partner integrations, including SAP, working from sparse legacy documentation. Both platforms ran side by side for months, through the busiest period of the year. Nothing switched over until the outputs matched.
First technical hire at a startup offering local mobile plans across many countries. I made the architecture decisions, designed the integration layer and built the Python gateway that put the carriers behind one API. I set the API standards, chose the stack for the move from MVP to services, joined pre-sales calls for customers with special requirements, and helped hire and guide the developers who joined after me. I also helped build an internal AI assistant used across engineering, support and sales.
"You've been an absolute powerhouse getting that done - impressive work!"
Theis Jensen, Co-founder, Telgea, on LinkedIn
I built an evaluation that ran sales-call transcripts through three Claude model tiers and had the strongest model grade the results. The mid tier matched the top tier, so the team got a repeatable way to choose models on cost.
How I work
I'm Danish, with a master's degree in Informatics from Aalborg University. For ten years I've been the person who makes systems talk to each other, as the founding engineer at startups and as a consultant placed with public-sector and enterprise clients.
Most of my projects started as a business problem rather than a spec: a regulator handling cases by hand, a startup that needed eight carriers behind one API. I'm comfortable starting there, before the problem is fully defined, and turning it into a clear scope and an architecture the business can make decisions on.
Contact
A few lines is enough, and I'll say whether I'm the right person for it.